Selena Deckelmann Emma, Inc - PostgreSQL ...

decubitisannouncerData Management

Nov 27, 2012 (4 years and 8 months ago)


Managing Terabytes
Selena Deckelmann
Emma, Inc -
PostgreSQL Global Development Group
Environment at Emma

1.6 TB, 1 cluster, Version 8.2 (RAID10)

1.1 TB, 2 clusters, Version 8.3 (RAID10)

8.4, 9.0 Dev

Putting 9.0 into production (May 2011)

pgpool, Redis, RabbitMQ, NFS
Other stats

daily peaks: 3000 commits per second

average writes: 4 MBps

average reads: 8 MBps

From benchmarks we’ve done, load is
pushing the limits of our hardware.
I say all of this with
Huge catalogs

409,994 tables

Minor mistake in parent table definitions

Parent table updates take 30 minutes
not null default nextval('important_sequence'::text)
not null default nextval('important_sequence'::regclass)
Huge catalogs

Bloat in the catalog

User-provoked ALTER TABLE

VACUUM FULL of catalog takes 2 hrs
Huge catalogs suck

9,019,868 total data points for table stats

4,550,770 total data points for index stats

Stats collection is slow
Disk Management


pg_tblspc (TABLESPACES)



wal for warm standby
Problems we worked through
with big schemas Postgres



System resource exhaustion

Minor upgrades

Major upgrades

Transaction wraparound
Bloat Causes

Frequent UPDATE patterns

Frequent DELETEs without VACUUM

a terabyte of dead tuples
schemaname, tablename, reltuples::bigint, relpages::bigint, otta,
ROUND(CASE WHEN otta=0 THEN 0.0 ELSE sml.relpages/otta::numeric END,1) AS tbloat,
CASE WHEN relpages < otta THEN 0 ELSE relpages::bigint - otta END AS wastedpages,
CASE WHEN relpages < otta THEN 0 ELSE bs*(sml.relpages-otta)::bigint END AS wastedbytes,
CASE WHEN relpages < otta THEN '0 bytes'::text ELSE (bs*(relpages-otta))::bigint || ' bytes' END AS wastedsize,
iname, ituples::bigint, ipages::bigint, iotta,
ROUND(CASE WHEN iotta=0 OR ipages=0 THEN 0.0 ELSE ipages/iotta::numeric END,1) AS ibloat,
CASE WHEN ipages < iotta THEN 0 ELSE ipages::bigint - iotta END AS wastedipages,
CASE WHEN ipages < iotta THEN 0 ELSE bs*(ipages-iotta) END AS wastedibytes,
CASE WHEN ipages < iotta THEN '0 bytes' ELSE (bs*(ipages-iotta))::bigint || ' bytes' END AS wastedisize
schemaname, tablename, cc.reltuples, cc.relpages, bs,
(CASE WHEN datahdr%ma=0 THEN ma ELSE datahdr%ma END))+nullhdr2+4))/(bs-20::float)) AS otta,
COALESCE(c2.relname,'?') AS iname, COALESCE(c2.reltuples,0) AS ituples, COALESCE(c2.relpages,0) AS ipages,
COALESCE(CEIL((c2.reltuples*(datahdr-12))/(bs-20::float)),0) AS iotta
(datawidth+(hdr+ma-(case when hdr%ma=0 THEN ma ELSE hdr%ma END)))::numeric AS datahdr,
(maxfracsum*(nullhdr+ma-(case when nullhdr%ma=0 THEN ma ELSE nullhdr%ma END))) AS nullhdr2
schemaname, tablename, hdr, ma, bs,
SUM((1-null_frac)*avg_width) AS datawidth,
MAX(null_frac) AS maxfracsum,
SELECT 1+count(*)/8
FROM pg_stats s2
WHERE null_frac<>0 AND s2.schemaname = s.schemaname AND s2.tablename = s.tablename
) AS nullhdr
FROM pg_stats s, (
(SELECT current_setting('block_size')::numeric) AS bs,
CASE WHEN substring(v,12,3) IN ('8.0','8.1','8.2') THEN 27 ELSE 23 END AS hdr,
CASE WHEN v ~ 'mingw32' THEN 8 ELSE 4 END AS ma
FROM (SELECT version() AS v) AS foo
) AS constants
GROUP BY 1,2,3,4,5
) AS foo
) AS rs
JOIN pg_class cc ON cc.relname = rs.tablename
JOIN pg_namespace nn ON cc.relnamespace = nn.oid AND nn.nspname = rs.schemaname AND nn.nspname <> 'information_schema'
LEFT JOIN pg_index i ON indrelid = cc.oid
LEFT JOIN pg_class c2 ON c2.oid = i.indexrelid
) AS sml
WHERE tablename = 'addr'
ORDER BY wastedbytes DESC LIMIT 1;
Fixing bloat

Wrote scripts to clean things up

VACUUM (for small amounts)


TRUNCATE (data loss!)

Or most extreme: DROP/CREATE

And then ran the scripts.

pg_dump takes longer and longer
     backup      |                duration                
 2009-­‐11-­‐22  |  02:44:36.821475
 2009-­‐11-­‐23  |  02:46:20.003507
 2009-­‐11-­‐24  |  02:47:06.260705
 2009-­‐12-­‐06  |  07:13:04.174964
 2009-­‐12-­‐13  |  05:00:01.082676
 2009-­‐12-­‐20  |  06:24:49.433043
 2009-­‐12-­‐27  |  05:35:20.551477
 2010-­‐01-­‐03  |  07:36:49.651492
 2010-­‐01-­‐10  |  05:55:02.396163
 2010-­‐01-­‐17  |  07:32:33.277559
 2010-­‐01-­‐24  |  06:22:46.522319
 2010-­‐01-­‐31  |  10:48:13.060888
 2010-­‐02-­‐07  |  21:21:47.77618
 2010-­‐02-­‐14  |  14:32:04.638267
 2010-­‐02-­‐21  |  11:34:42.353244
 2010-­‐02-­‐28  |  11:13:02.102345

pg_dump fails

patching pg_dump for SELECT ... LIMIT

Crank down shared_buffers

Install 32-bit Postgres and libraries on a 64-bit system.
Install 64-bit Postgres/libs of the same version.
Copy “hot backup” from 32-bit sys over to 64-bit sys.
Run pg_dump from 64-bit version on 32-bit Postgres.

Warm standby is not a backup

Hot backup instances

“You don’t have valid backups, you have
valid restores.” (thanks sarahnovotny)

Necessity is the mother of invention...
Ship WAL from Solaris x86 - Linux
It did work!
Running out of inodes

UFS on Solaris
“The only way to add more inodes to a UFS
filesystem is:
1. destroy the filesystem and create a new
filesystem with a higher inode density
2. enlarge the filesystem - growfs man page”

Solution 0: Delete files.

Solution 1: Sharding and bigger FS on Linux

Solution 2: ext4 (soon!)
Running out of
available file descriptors

Too many open files by the database

Pooling - pgpool or pgbouncer?
Minor upgrades

Stop/start database

CHECKPOINT() before shutdown
Major Version upgrades

Too much downtime to dump/restore

Write tools to migrate data

Trigger-based replication

wraparound avoidance

autovacuum triggers are too small

Watch age(datfrozenxid)

Increase autovacuum_freeze_max_age

We’re hiring! -

Emma’s Tech Blog:

My blog: